Pipeline Gpu Kernel

Software pipelining (double-buffer) on tiled GPU kernel → overlap global mem loads w/ Tensor Core compute. Covers prologue/loop/epilogue restructure, LDG-reg vs cp.async (LDGSTS) variant pick by compute/load ratio, smem budget vs arch occupancy cliff, SASS-level overlap verify.

pjt222 Updated 44 repo stars

File contents

pjt222/agent-almanac commit bb9e4f5b18

Frequently asked questions

npx get-skill add pjt222/pipeline-gpu-kernel-10